Canadian news publishers want Meta to scrap new terms of service over AI training concerns

A group representing top Canadian news publishers is calling on Meta Platforms Inc. to scrap controversial new commercial terms of service over concerns that the language used could permit the American tech giant to scrape companies’ content to train its artificial intelligence models.

Last week, Meta temporarily updated the terms of service for its business accounts before delaying the implementation after backlash from media and digital rights groups in the United States and Europe.

The new language said that users of its products authorized Meta to employ tools such as web crawlers to scrape content from sites associated to the users to “develop, improve, research, train … Meta company products, service and features provided under these commercial terms and any other applicable supplemental terms.”

Meta owns social media and messaging apps Facebook, Instagram and WhatsApp. The company has also spent billions on its AI efforts in a bid to catch up with frontier labs such as OpenAI Group PBC and Anthropic PBC, and said this week its spending will grow to US$825 billion, with much of it directed toward AI computing power and infrastructure.

The company has since removed the new language and pushed back the proposed implementation date, first to Oct. 30 and then to Nov. 20.

An advocacy group representing Canadian largest publishers said the changes are not enough to assuage their worries, and want Ottawa to intervene if Meta keeps the new terms intact.

“We are still very much concerned,” said Paul Deegan, president and chief executive officer of News Media Canada, a trade industry body representing publishers including The Globe and Mail Inc., Torstar Corp. and Postmedia Network Canada Corp., which publishes the Financial Post.

“The revised terms don’t actually spell out that Meta will abide by the restrictions, nor do they promise that Meta will limit scraping to identified bots. They do not limit how content collected by that crawler may be used,” he said.

Deegan said that the impact of Meta’s terms extends beyond news publishers.

“Every publicly-hosted Canadian word, song and image created is up for grabs if Meta doesn’t reverse course or if the government doesn’t intervene,” he said, adding that News Media Canada has taken the publishers’ concerns to the Canadian government.

In response to the backlash, Meta told multiple media outlets last week its new terms don’t change how the company uses web crawlers.

“As always, content owners can use robots.txt to signal their preferences,” referring to a text code that communicates to web crawlers which parts of a website it can access and which parts are private.

“This is not about scraping news sites and these updates don’t give us new rights that supersede those preferences,” the company said.

But Deegan said that blocking crawlers is not a practical option for publishers, who need to allow Meta’s ad crawlers access in order for their advertising to function.

Antitrust bodies in Europe and the United Kingdom said they have received complaints from publishers and digital advertisers over Meta’s updated terms. The U.K.’s competition watchdog met with Meta this week, while the EU’s antitrust regulator said it welcomes the company’s postponement of the implementation date.

The News/Media Alliance (NMA), a trade body representing over 2,000 newspaper, magazine and digital publishers in the U.S., argued the terms “grant Meta extraordinary and disproportionate privileges to exploit the content of its websites for free,” leaving publishers no way of opting out unless they leave Meta’s platforms and its “massive audience,” the group said in a letter to the company dated Oct. 22.

It urged Meta to withdraw the terms and hold discussions around future changes with the group and other publishers.

Meta said in an emailed statement on Friday that it delayed the implementation date of its new service terms to give it more time to explain how it uses web crawlers and “what’s changing and what’s not.”

Canadian news publishers are already engaged in legal battles with U.S. AI giants over copyright issues.

Most recently, Postmedia, The Globe and Mail, Torstar and the Canadian Press filed a lawsuit against Anthropic in Canadian Court on Oct. 6, arguing that the AI giant breached copyright and illegally used their content to train its AI models. Those allegations have not been proven in court.